Genetic Algorithms in path planning Mansoureh Ziaei Intelligent - PowerPoint PPT Presentation

Genetic Algorithms in path planning Mansoureh Ziaei Intelligent Robotics Seminar, Group TAMS, University of Hamburg

Agenda I. Concepts of Genetic Algorithms A. Parameters B. Limitations C. Variants D. Problem domains II. Path planning [http://imgkid.com] 2

Darwin’s theory of evolution Organisms change over time As a result of changes in heritable physical or behavioral traits To better adapt to its environment To survive and have more offspring Microevolution Macroevolution [http://www.livescience.com] 3

Why evolution as an inspiration for computational problems? Searching through a huge number of possibilities Effective use of parallelism Adaptive computer programs Innovative computer programs Complex solutions 4

Genetic Algorithms Invented by John Holland in 1960s Goal: importing natural adaption into computer systems rather than solving specific problems [http://evolutionarysystemsbiology.org] 5

How? - Genetic representation of the solution domain - Fitness function to evaluate the solution domain - Selection mechanisms - Crossover - Mutation 6

[http://www.edc.ncl.ac.uk] GAs parameters 7

- Fitness function evaluation - Scaling with complexity - Stop criterion Limitations - Local optima - Dynamic data sets - Right/wrong measure - Speed of convergence 8

- Chromosome representation Variants - Elitism - Parallel implementations - Adaptive GAs 9

- Timetabling and scheduling Problem - Engineering - Global optimization problem domains - Pattern /speech recognition - Training neural networks - Path planning - ... 10

Path planning 11

Autonomous robot navigation - Explore an environment independent of human presence or intervention - Uncertain environment - Generate a feasible path and optimize it in world space - Global planning vs. local planning [2] 12

Using genetic algorithms - Value-encoded scheme - Path fitness based on path length and path feasibility [2] [2] 13

Applied elements Fitness function: [4] Rank selection Single cross point Mutation Elitism [4] 14

Thanks! Questions? Bibliography: 1. Melanie, Mitchell. "An introduction to genetic algorithms." Cambridge, Massachusetts London, England, Fifth printing 3 (1999). 2. Manikas, W., Kaveh Ashenayi, and RogerL Wainwright. "Genetic algorithms for autonomous robot navigation." Instrumentation & Measurement Magazine, IEEE 10.6 (2007): 26-31. 3. Alvarez, Alberto, Andrea Caiti, and Reiner Onken. "Evolutionary path planning for autonomous underwater vehicles in a variable ocean." Oceanic Engineering, IEEE Journal of 29.2 (2004): 418-429. 4. Sedighi, Kamran H., et al. "Autonomous local path planning for a mobile robot using a genetic algorithm." Evolutionary Computation, 2004. CEC2004. Congress on . Vol. 2. IEEE, 2004. 15